JupyterHub user environments have wrong environment
Using a conda environment inside a jupyter notebook/lab does not source the full shell environment of that user. It is easy to test.
- create a conda environment or any other.
- add the kernel to jupyter
- load that kernel
- run:
!echo $PATH
It is clear then, that the environment is very limited and other variables are missing too. This is not usually a problem, but on JET this results in a lot of missing paths.
This is related to the issue, that Teaching Kernels do not have the correct path set when used in TeachingHub.
e.g.: modell2022 showing:
{
"argv": [
"/usr/local/bin/kernel-launcher",
"",
"/home/swd/jupyterhub/alpha/envs/modell2022/bin/python",
"-m",
"ipykernel_launcher",
"-f",
"{connection_file}"
],
"display_name": "MSC-Modelling-WS2022",
"language": "python"
}